What Is Acoustic Monitoring and How Is It Used?

Acoustic monitoring is the practice of using microphones, hydrophones, or vibration sensors to capture sound and then analyzing those recordings to learn something about the source. It spans an enormous range of applications: ecologists leave small recorders in forests to track bird populations, engineers attach sensors to wind-turbine gearboxes to catch early signs of bearing damage, and physicians are developing wearable stethoscopes that continuously listen to a patient’s lungs. The unifying idea is that sound carries information, and with the right hardware and analysis, you can extract that information remotely, continuously, and often at a fraction of the cost of sending a human to look.

Passive Versus Active Listening

In ecology, the term most people encounter is passive acoustic monitoring, or PAM. “Passive” means the recorder simply listens to whatever sounds the environment produces. It does not emit a signal of its own. A PAM device left in a wetland captures frog choruses, insect stridulations, wind, rain, and the occasional passing vehicle without disturbing any of them. By contrast, active acoustic methods do exist: sonar pings used to map the seafloor or ultrasonic pulses bounced off a material to check for internal cracks. But the passive approach has become the dominant paradigm for biodiversity research because it is cheap, non-invasive, and can run unattended for weeks or months.

A study comparing passive recorders with traditional human surveys in tropical savanna wetlands illustrates why. The passive devices detected all frog species found by human surveyors and consistently found more: roughly 14 species per site versus about 8 with active human searches. The advantage grew as the recorders covered a wider slice of the day and season, picking up species that call only at odd hours or during brief windows.

¹

What the Hardware Looks Like

For much of the twentieth century, acoustic fieldwork meant heavy reel-to-reel tape recorders and a researcher holding a parabolic dish. That changed dramatically with the arrival of small, programmable digital recorders. The most widely adopted is AudioMoth, an open-source device about the size of a credit card with a build cost of roughly $43 per unit. It can sample at rates up to 384 kHz, high enough to capture the ultrasonic calls of bats, and runs on three AA batteries.

²

Because AudioMoth is open-source, researchers have adapted it freely. One offshoot, HydroMoth, wraps similar electronics in a waterproof housing for underwater recording at a fraction of the price of traditional hydrophones.

³

Detection algorithms can also be loaded directly onto the device so it records only when it hears something of interest, dramatically cutting power draw and storage needs.

⁴

Tracking Birds, Bats, and Other Terrestrial Wildlife

Terrestrial biodiversity surveys are the most common use of PAM. Arrays of recorders deployed across a landscape can tell you not just which species are present but how active they are and how their behavior changes with habitat quality. One study using automated PAM around hedgerows of different ages in the UK found that bird and bat vocal activity was significantly higher around older, established hedgerows compared to newly planted ones, with young hedgerows (a few years old) showing the strongest gains. That kind of finding helps land managers understand how quickly habitat restoration translates into measurable wildlife recovery.

⁵

Bat monitoring is a particularly strong use case because bats are nocturnal and echolocate at frequencies humans cannot hear. Autonomous recording units pick up these calls reliably, though automated classification software does occasionally confuse species that sound alike. A study of bat surveys on military lands in Colorado estimated that roughly 1 to 5 percent of detections of certain bat species were misclassified, a useful reminder that the technology works well but is not perfect.

⁶

Acoustic monitoring even extends to animals whose calls sit below the range of human hearing. African elephants produce low-frequency rumbles, some with fundamental frequencies between 10 and 25 Hz, that travel long distances through both air and ground. Researchers at a zoo near Frankfurt recorded over 1,350 rumbles using co-located seismic and infrasound sensors, demonstrating that you can track elephant vocal behavior non-invasively with ground-mounted equipment.

⁷

Listening to the Ocean

Underwater, sound travels farther and faster than in air, making acoustic monitoring especially powerful for marine ecosystems. Coral reefs are surprisingly noisy places. Healthy reefs hum with the clicks, grunts, and pops of fish and invertebrates, while degraded reefs go comparatively quiet. This difference has turned reef soundscapes into a potential diagnostic tool. A study comparing healthy, restored, and degraded coral reefs found that healthy and restored reefs had a similar diversity of biological sounds, significantly higher than degraded sites.

⁸

To standardize what “healthy-sounding” means, researchers have developed acoustic indices. The Acoustic Complexity Index (ACI), for instance, measures how variable the sound is across frequencies and time, capturing the richness of a soundscape. On one tropical reef, a low-frequency version of ACI was influenced by fish knocks, herbivorous feeding sounds, and even fishing pressure, explaining about half the variation in the index.

⁹

In snapping-shrimp-dense waters like those around Singapore, researchers have explored snap rates as an additional metric, since shrimp activity reflects local conditions. Previous work has found positive correlations between sound-pressure levels and live coral cover, and between ACI and fish diversity.

¹⁰

That said, not every acoustic metric tells the same story. The coral restoration study noted that some automated indices replicated the pattern found by manually counting distinct biological sounds, but others did not. This is still a maturing field, and choosing the right metric for the right question matters.

⁸

Freshwater Environments

Rivers, lakes, and ponds have received far less acoustic attention than forests or oceans, but that is changing. Freshwater environments host a surprising range of sound-producing organisms: fish, frogs, and aquatic insects including beetles, water bugs, dragonfly larvae, and caddisfly larvae.

¹¹

Physical processes like water flow and sediment transport add their own acoustic signatures, and human activities on or near the water contribute too. PAM in freshwater has already revealed unexpected acoustic diversity, suggesting it could become a practical tool for assessing river and lake health the way it is starting to work for coral reefs.

¹²

Soundscape Ecology and the Bigger Picture

Rather than focusing on one species at a time, some researchers treat the entire soundscape as a unit of study. The field of soundscape ecology divides environmental sound into three broad categories: biological sounds (biophony), geophysical sounds like wind and rain (geophony), and human-produced sounds (anthrophony). By analyzing the balance among these three, you can draw inferences about ecosystem health without needing to identify every individual species.

¹³

This approach has practical appeal because species-level identification from recordings is labor-intensive and error-prone. A soundscape index can flag that something changed, acting like an early-warning system, even if you do not know exactly which species drove the change. Conservationists can then follow up with targeted surveys in areas where the acoustic signature has shifted.

How AI Is Changing the Analysis

A single AudioMoth recording for a month generates terabytes of audio. No human team can listen to all of it, so machine learning has become essential. Deep-learning models trained on spectrograms (visual representations of sound) can now identify bird species from their songs with accuracy above 96 percent on multiple datasets.

¹⁴

Noise is the practical bottleneck. Field recordings come packed with wind, rain, traffic rumble, and other interference that can fool classifiers. Dedicated denoising networks can boost the signal-to-noise ratio of bird sounds by more than 10 decibels in most cases, with some scenarios seeing improvements of over 35 decibels.

¹⁵

That kind of preprocessing makes the downstream species classifiers far more reliable, especially in noisy urban or roadside environments where conservation questions are often most urgent.

Noise Pollution Research

Acoustic monitoring is not just a way to study wildlife; it is also how we study what noise pollution does to wildlife. By recording animals in areas with different levels of human noise and comparing their calls, researchers have found that terrestrial species, especially birds, tend to raise the minimum frequency of their calls in response to noise pollution. They do not generally change how loud or how long they call; instead, they shift the pitch upward, presumably to be heard above the low-frequency rumble of traffic or machinery.

¹⁶

The consequences extend beyond communication. A systematic review of noise effects on animal behavior found altered movement patterns, reduced foraging, disrupted reproduction and offspring care, and escape reactions in marine animals. Birds showed changes in foraging and nesting, while laboratory studies on small mammals linked chronic noise exposure to spatial-cognitive impairments and memory loss.

¹⁷

Without PAM and related acoustic tools, these patterns would be nearly impossible to quantify at scale.

Industrial Condition Monitoring

Step away from the natural world and acoustic monitoring looks quite different. In industrial settings, the relevant sounds are stress waves: tiny, high-frequency elastic pulses released when materials crack, deform, or rub against each other under load. Sensors that detect these acoustic emissions are attached to machinery, pipelines, or structures to catch problems before they cause failures.

¹⁸

Wind-turbine gearboxes are a good example of why this matters. Bearing failures in wind turbines lead to long downtimes and expensive repairs. Acoustic-emission-based monitoring applied to roller bearings detected surface damage at least 8 hours (about 4 percent of time-to-failure) earlier than traditional vibration sensors. More interesting, the system picked up significant deviations from baseline roughly 130 hours, or about half the remaining life, before test stop, possibly catching sub-surface damage that had not yet reached the bearing surface.

¹⁹

Structural Health Monitoring

The same principle applies to bridges, buildings, and pressure vessels. Acoustic emission sensors glued to a steel beam can track fatigue cracks in real time. As a crack grows, the acoustic signals it produces change in predictable ways: energy, event count, and signal duration all rise, and they track each other closely enough to serve as reliable indicators of how fast the crack is spreading.

²⁰

This kind of monitoring can supplement or eventually replace scheduled inspections, catching damage that might not be visible on the surface until it is already serious.

Medical Auscultation Goes Wearable

Listening to the body with a stethoscope is one of the oldest forms of acoustic monitoring. Researchers are now miniaturizing and automating the process. Soft, wearable stethoscope patches can sit on a patient’s chest and continuously stream heart and lung sounds wirelessly to a phone or clinical dashboard. These devices can separate cardiac from respiratory signals even in noisy environments and flag abnormal patterns in real time.

²¹

One system, designed as a lung-sound monitoring patch, pairs a wearable sensor with a mobile app that classifies normal versus adventitious (abnormal) breathing sounds by comparing their unique acoustic characteristics. It can distinguish heart sounds from breath sounds and filter out external noise, opening up possibilities for continuous home monitoring of people with chronic respiratory conditions.

²²

Smart Cities and Urban Noise Mapping

Urban planners have adopted acoustic monitoring to map noise pollution across cities. Networks of wireless acoustic sensors can measure sound-pressure levels continuously and feed the data into models that guide traffic management, building codes, and green-space placement. The concept is sometimes framed as part of the Internet of Things infrastructure for smart cities.

²³

In Aveiro, Portugal, a project deployed nine noise sensors across the city, each recording continuous sound-pressure levels at five-minute intervals. The resulting data, combined with traffic models, let researchers map noise exposure with far greater spatial and temporal resolution than the occasional spot-check survey that cities traditionally rely on.

²⁴

Volcanoes and Infrasound

Some of the most dramatic acoustic monitoring happens at frequencies well below what any animal can hear. Explosive volcanic eruptions generate infrasound, pressure waves with frequencies below 20 Hz that can travel thousands of kilometers. The International Monitoring System, originally designed to detect nuclear tests, maintains a global network of infrasound arrays that also picks up volcanic activity. Researchers analyzing data from a station in Antarctica, over 1,600 km away, identified infrasound signals from episodes of persistent eruptive activity at Mount Michael in the South Sandwich Islands across a 16-year window. The signal amplitudes were consistent with moderate explosive eruptions, providing evidence of activity at a remote volcano that is difficult to observe any other way.

²⁵

Privacy and the Human Voice Problem

When you leave a microphone in the field to record frogs and birds, it will occasionally pick up people talking. This creates a tension between ecological data collection and individual privacy. Eco-acoustic monitoring is increasingly deployed at huge scales, yet relatively little thought has historically been given to the privacy implications of inadvertently recorded human speech.

²⁶

Automated speech-detection algorithms can now flag and strip segments of audio that contain human voices before the data are shared or archived. The same detection capability has a bonus scientific use: quantifying the level of human disturbance at a site by measuring how often speech is present. Still, as monitoring networks grow denser and recorders become cheaper, the ethics of ambient listening in public and semi-public spaces will need ongoing attention from both researchers and regulators. Some jurisdictions already treat ambient audio recording differently from, say, camera-trap photography, and the legal landscape is patchy enough that project-by-project review remains the safest approach.